All units
4Foundation Module · Unit 4 of 6
Risks of AI in Clinical Practice
Examines how AI fails in clinical practice — hallucination, bias, automation bias, error propagation from a single mistake to a system failure — and identifies where AI use becomes genuinely dangerous.
5 recorded lectures 5 knowledge checks
What you will be able to do
On completing this unit.
- Define an AI hallucination, explain why it is a predictable failure mode rather than a rare anomaly, and apply strategies for detecting one.
- Trace bias from its sources in training data through to harm at the bedside, particularly for patients from underrepresented populations.
- Recognise automation bias as a normal cognitive tendency rather than a personal failing, identify the situations that trigger it, and apply countermeasures that keep cognitive ownership with you.
- Follow how a single error propagates through copy-forward documentation, handovers and team workflows, and intervene early enough to break the chain.
- Stratify clinical situations by risk — complexity, atypical presentation, time pressure, poor input quality — and set conservative boundaries for where AI is not used at all.
Judgment first. AI second.
Lectures in this unit
- 4AAI Hallucinations (When Fluent Answers Are Wrong)
- 4BBias in Medical AI (Data, Distortion, and Risk)
- 4CAutomation Bias (Over-Trust and Cognitive Offloading)
- 4DError Propagation (From Single Mistake to System Failure)
- 4EHigh-Risk AI Use (When AI Becomes Dangerous)
This unit is part of the Foundation Module
Six units, 30 recorded lectures, about four hours. The Foundation Module is the entry requirement for every advanced module that follows.